City University of New York (CUNY) City University of New York (CUNY)

CUNY Academic Works CUNY Academic Works

Publications and Research City College of New York

2015

Population Genomics Reveals Seahorses (Hippocampus erectus) Population Genomics Reveals Seahorses (Hippocampus erectus)

of the Western Mid-Atlantic Coast to Be Residents Rather than of the Western Mid-Atlantic Coast to Be Residents Rather than

Vagrants Vagrants

J. T. Boehm CUNY City College

John Waldman CUNY Queens College

John D. Robinson Marine Resources Research Institute

Michael J. Hickerson CUNY City College

How does access to this work benefit you? Let us know!

More information about this work at: https://academicworks.cuny.edu/cc_pubs/17

Discover additional works at: https://academicworks.cuny.edu

This work is made publicly available by the City University of New York (CUNY). Contact: AcademicWorks@cuny.edu

RESEARCH ARTICLE

Population Genomics Reveals Seahorses(Hippocampus erectus) of the Western Mid-Atlantic Coast to Be Residents Rather thanVagrantsJ. T. Boehm1,3*, JohnWaldman2,3, John D. Robinson4, Michael J. Hickerson1,3

1 Department of Biology, City College of New York, 160 Convent Ave., New York, New York, 10031, UnitedStates of America, 2 Biology Department, Queens College, City University of New York, 65-30 KissenaBlvd., Queens, New York, 11367-1597, United States of America, 3 Subprogram in Ecology, Evolution andBehavior, The Graduate Center of the City University of New York, 365 5th Ave, New York, New York, 10016,United States of America, 4 South Carolina Department of Natural Resources, Marine Resources ResearchInstitute, 217 Fort Johnson Rd., Charleston, South Carolina, 29412, United States of America

* jtboehmjr@gmail.com

AbstractUnderstanding population structure and areas of demographic persistence and transients is

critical for effective species management. However, direct observational evidence to ad-

dress the geographic scale and delineation of ephemeral or persistent populations for many

marine fishes is limited. The Lined seahorse (Hippocampus erectus) can be commonly

found in three western Atlantic zoogeographic provinces, though inhabitants of the temper-

ate northern Virginia Province are often considered tropical vagrants that only arrive during

warm seasons from the southern provinces and perish as temperatures decline. Although

genetics can locate regions of historical population persistence and isolation, previous

evidence of Virginia Province persistence is only provisional due to limited genetic sampling

(i.e., mitochondrial DNA and five nuclear loci). To test alternative hypotheses of historical

persistence versus the ephemerality of a northern Virginia Province population we used a

RADseq generated dataset consisting of 11,708 single nucleotide polymorphisms (SNP)

sampled from individuals collected from the eastern Gulf of Mexico to Long Island, NY. Con-

cordant results from genomic analyses all infer three genetically divergent subpopulations,

and strongly support Virginia Province inhabitants as a genetically diverged and a historical-

ly persistent ancestral gene pool. These results suggest that individuals that emerge in

coastal areas during the warm season can be considered “local” and supports offshore mi-

gration during the colder months. This research demonstrates how a large number of genes

sampled across a geographical range can capture the diversity of coalescent histories

(across loci) while inferring population history. Moreover, these results clearly demonstrate

the utility of population genomic data to infer peripheral subpopulation persistence in diffi-

cult-to-observe species.

PLOS ONE | DOI:10.1371/journal.pone.0116219 January 28, 2015 1 / 12

OPEN ACCESS

Citation: Boehm JT, Waldman J, Robinson JD,Hickerson MJ (2015) Population Genomics RevealsSeahorses (Hippocampus erectus) of the WesternMid-Atlantic Coast to Be Residents Rather thanVagrants. PLoS ONE 10(1): e0116219. doi:10.1371/journal.pone.0116219

Academic Editor: Matthias Stöck, Leibniz-Institute ofFreshwater Ecology and Inland Fisheries, GERMANY

Received: July 25, 2014

Accepted: October 20, 2014

Published: January 28, 2015

Copyright: This is an open access article, free of allcopyright, and may be freely reproduced, distributed,transmitted, modified, built upon, or otherwise usedby anyone for any lawful purpose. The work is madeavailable under the Creative Commons CC0 publicdomain dedication.

Data Availability Statement: Illumina generatedsequence reads for each individual are available atNCBI Sequence Read Archive (SRA): SRP048776.Additional data are available from Dryad(doi:10.5061/dryad.qc2qq).

Funding: The National Science Foundation (40C33-00-01) supported the dissertational research ofJ. T. B. The funders had no role in study design, datacollection and analysis, decision to publish, orpreparation of the manuscript.

IntroductionIn warmer seasons, the waters lining the concrete bulkheads, wooden piers, estuaries, andsandy beaches of the temperate Northeastern United State’s mid-Atlantic coast become hometo numerous tropical fish species [1,2]. Over a century of research has cataloged the immigra-tion of tropical vagrants or “strays” to these coastal mid-Atlantic waters. The majority of theseindividuals arrive due to passive planktonic dispersal in summer months, transported by oceancurrents that circle north off the warm water mass of the Gulf Stream as it deflects northeastfrom U.S. towards Europe at roughly 35°N latitude [3,4]. This phenomenon positions CapeHatteras as a delineation point between the zoogeographic Virginia and Carolina Provinces,each defined by distinct faunal endemism and unique macroclimatic conditions (Fig. 1) [5,6].Following this observation, studies of species found in both provinces suggest that Cape Hat-teras acts as a “barrier” where intraspecific gene flow is reduced between provinces or alterna-tively acts as a northern latitudinal limit during the winter for species without cold thermaltolerance [6,7]. In this latter case, more sedentary tropical species that passively drift into thetemperate Virginia Province during warmer months locally perish after cold wintertemperatures advance.

Though many fishes exhibit wide thermal tolerance, ascertaining the true range of marinespecies can be challenging due to factors that include patchy distributions, cyclical populationsizes, and seasonal movement patterns [7,8]. One species often associated with tropical va-grants in the Virginia Province is the Lined seahorse,Hippocampus erectus [3]. Its status as apersistent independent gene pool (i.e., subpopulation) is uncertain primarily due to its near-shore absence during cold winter months and a scarcity of direct winter observations of indi-viduals. H. erectus is commonly found in coastal zones in three zoogeographic provinces:Caribbean (tropical), Carolina (warm-temperate) and Virginia (temperate) (Fig. 1). Some re-searchers suggest that long-distance rafting carries migrants northward to temporarily inhabitthe Virginia Province as temperatures warm [3], a prediction supported by substantial observa-tional evidence of long-distance rafting migration throughout its range [9,10]. In contrast,other researchers suggest that localized active dispersal directed toward offshore migration forthermal refuge in continental shelf waters during late fall accounts for its winter absence [11].This hypothesis of seasonal localized migration is partially supported by the observation of in-shore colonization ofH. erectus as temperatures warm in April to June, characteristic of mosttemperately adapted fishes [3], and earlier than the July to September arrival typical for the ma-jority of tropical strays [1,12].

Ecologists and evolutionary biologists often focus on questions at different temporal scales,but both fields are increasingly making use of genetic data to test hypotheses about populationhistory, estimate the movement of individuals between local populations, and characterize thespatial distribution of genetic variation for effective species management [8,13,14]. One exam-ple is the use of genetic data to examine source-sink dynamics [15,16]. True sink populations,even if annually persistent, require continual immigration from source populations and are ex-pected to exhibit genetic homogeneity with source populations or heterogeneity reflecting mul-tiple sources of immigrants, while over time independent breeding subpopulations throughrandom (genetic drift) or deterministic (natural selection) processes will exhibit distinct geneticdivergence [13]. A number of studies have examined the biogeography and genetic divergenceofHippocampus species [17]. Most of this research has focused on Indo-Pacific species withgenetic variation ranging in spatial scales from among localized South African estuaries [18], towidespread species complexes associated with rafting driven colonization [19], and differinglevels of intraspecific divergence attributed to both ecological traits and biogeographic divides[20,21].

Seahorses of the Temperate Virginia Province

PLOS ONE | DOI:10.1371/journal.pone.0116219 January 28, 2015 2 / 12

Competing Interests: The authors have declaredthat no competing interests exist.

Here we test whether the presence of H. erectus individuals north of Cape Hatteras are theresult of an ephemeral deme that is seasonally replenished from demographically persistentsouthern populations (H1; Hypothesis 1), or in contrast, there are persistent and isolatedpopulations on either side of Cape Hatteras (H2; Hypothesis 2). A previous study of theH. erectus complex utilized mitochondrial DNA and five more slowly evolving nuclear lociacross many individuals (n = 115), yet rejected H2 in favor of H1, with little divergence and ev-idence of isolation across Cape Hatteras [22]. Now, with the decreasing cost of high-through-put sequencing, data can be sampled from across the autosomal genome to account forvariations in mutation, coalescent history, and recombination, thereby facilitating a view of thecomplexity of a species evolutionary history with the potential to infer more recent divergenceand/or populations differentiating in the presence of gene flow [23].

To date, genome wide single nucleotide polymorphism (SNP) datasets generated byrestriction site associated DNA sequencing (i.e., RADseq) have been utilized to study severalfish species. Examples utilizing RADseq datasets include the support of cryptic differentiationbetween populations of the Baltic Sea herring (Clupea herangus) [24], the detection of hybridindividuals between trout species [25], genetic divergence of various stickleback populations[26–28], and robust phylogenetic resolution between African cichlid species [29]. To test theaforementioned competing hypotheses H1 and H2, we generated a genomic RADseq datasetconsisting of 11,708 SNPs across individuals of H. erectus from the eastern Gulf of Mexico toLong Island, NY (Fig. 1).

Although we base our inference from only 4–9 individuals per each of the three zoogeo-graphic provinces (total individuals; n = 23), data from large numbers of unlinked loci allowhighly resolved inference even with few individuals [30–32]. Moreover, given that outbred dip-loid genomes are comprised of recombining segments of DNA inherited from large pools ofancestors [33], genome-level datasets should capture the diversity of coalescent histories(across loci) that reflects population history, such that information comes more from the

Figure 1. Map of zoogeographic provinces, collection sites, and temperature variance.Contrastingocean minimum sea surface temperatures across zoogeographic provinces: generated in ARCGIS v.9.3using the Bio-ORACLE long-term climatic dataset [67]. Collection sites from the northeastern Gulf of Mexicoto New York State indicated by diamonds: Apalachicola, Tampa Bay-Charlotte Harbor, Florida Keys, IndianRiver Lagoon and Jacksonville, FL., Chesapeake Bay, New Jersey-New York.

doi:10.1371/journal.pone.0116219.g001

Seahorses of the Temperate Virginia Province

PLOS ONE | DOI:10.1371/journal.pone.0116219 January 28, 2015 3 / 12

number of loci sampled through the genome than from numbers of individuals per samplinglocality [34,35].

Methods and Materials

Sampling and bioinformaticsSamples of H. erectus ranged from the eastern Gulf of Mexico to New York State (n = 23).Samples were collected from 2009–2013 from the following locations: Apalachicola, FL, TampaBay, FL, Charlotte Harbor, FL, the Florida Keys, Jacksonville, FL, and Indian River Lagoon, FL,Chesapeake Bay, New Jersey, the Hudson River and Long Island, NY. The specimens collectedin this study were carried out in accordance and approval of the Queens College InstitutionalAnimal Care and Use Committee (IACUC) (Permit # 137), which approved all aspects of spec-imen use in this study. Domestic fishing ofHippocampus is neither under direct regulationwithin the United States nor under species protection and no specific permissions were re-quired for these locations; however we collaborated with the following authorities for samples,and if standard collection permits were required they were issued for each collection location.The Florida specimens used in our study were collected under the authority of the Florida FishandWildlife (FFW) as part of the FFW: Southeast Area Monitoring and Assessment Program.Samples from the Chesapeake Bay were collected in collaboration with the Virginia Institute ofMarine Science (VIMS), which is authorized to collect any fishes necessary for research underthe Code of Virginia. Lastly, samples collected in New Jersey and New York were collected incollaboration with Rutgers University under the New Jersey Department of EnvironmentalProtection and the New York State Department of Environmental Conservation (DEC) SpecialLicensing Unit, License No. 1638 with additional samples collected under DEC License No.1405.

Sequenced samples were randomly chosen from a large number of individuals (n>100)over multiple collection years to ensure genomic similarity was not the result of non-independent relatedness. Total Genomic DNA was extracted using Puregene extraction(Qiagen) from tail muscle tissue and treated with RNAase A following standard protocols.Genomic DNA quality was checked on an agarose gel to ensure that the majority of DNA was>10,000bp and equalized to 30 ng/uL using Qubit Fluormetric Quantitation (Invitrogen). Li-brary construction and restriction site associated DNA sequencing (RADseq) protocol fol-lowed [36,37]. Floragenex carried out library preparation and sequencing. Genomic DNArestriction digestion utilized the Sbfl enzyme and individual sequence adapters and barcodeidentifiers were ligated to genomic DNA prior to sequencing on the Illumina HiSeq platform.All sequences from cut sites resulted in single-end reads, which were demultiplexed andtrimmed of adapters to 90bp fragment lengths.

Total reads per individual ranged from 1,264,862–4,736,299. The individual with the largestnumber of reads was processed to construct a de novo pseudo reference genome, and reads foreach individual were aligned using BOWTIE [38]. SAMTOOLS algorithms [39] and customFloragenex perl scripts were used to detect SNPs and call genotypes. SNP datasets were format-ted in the variant call format (vcf) [40]. Initial genotyping required a minimum Phred qualityscore of 15, a minimum of 4× sequence coverage, with a minimum of 65% of individuals geno-typed. Additional filtering was applied using R v.3 [41] to ensure a Phred score equal to a hardcutoff of q = 20 (base call accuracy lower than 99%). To reduce the inclusion of false SNP dis-covery due to paralogous sequences or low quality genotype calls, vcftools was utilized to re-move any sites with a minimum depth of 8× sequence coverage and maximum depthcalculated in R based on the mean depth + 1.5 standards deviation (= 295) across all sites. Thefinal datasets resulted in a bi-allelic matrix of 11,708 genotypes (5777 90bp sequences) across

Seahorses of the Temperate Virginia Province

PLOS ONE | DOI:10.1371/journal.pone.0116219 January 28, 2015 4 / 12

individuals at all sites. For details on per individual raw reads, filtered, and analyzed reads seeS1 Table.

Population genomic analysesA principal components analysis (PCA) was implemented to determine if sampled individualsreflect a history of differentiated populations by outputting individual coordinates along axesof genetic variation within a statistical framework [42] that correspond to the first two principlecomponents in Fig. 2B. To further aide in assigning individuals to differentiated populationsby inferring ancestry coefficients representing the proportions of each individual’s genome thatoriginated from a specified number of ancestral gene pools (K) we used the program sNMF[43]. The program sNMF estimates individual ancestry and population clustering by utilizing asparse non-negative matrix factorization algorithm (sNMF) to compute least-squares estimates

Figure 2. Genomic variation across individuals and subpopulations. (a) Treemix population tree with branch lengths scaled to the amount of genetic driftbetween regions and inferred proportion of genetic admixture (m = 2) between southern and northern regions represented by arrows. Dotted lines do notrepresent branch length. (b) Principle component analysis. Black circles = Chesapeake Bay-New York, dark grey circles = Florida Atlantic coast, and lightgrey circles = Gulf of Mexico-Florida Keys. Pie diagrams (a) represent ancestry coefficient proportions derived from the sNMF ancestry plot (c). Each line ofthe sNMF plot represents one individual.

doi:10.1371/journal.pone.0116219.g002

Seahorses of the Temperate Virginia Province

PLOS ONE | DOI:10.1371/journal.pone.0116219 January 28, 2015 5 / 12

of ancestry coefficients. This software is capable of efficiently analyzing large bi-allelic datasetswithout loss of accuracy when compared with more commonly utilized programs STRUC-TURE [44] and ADMIXTURE [45] that use the same underlying model to infer ancestry coeffi-cients. However, in contrast to the aforementioned programs, sNMF has significantly bettercomputational efficiency and is robust to many of the demographic assumptions of Hardy-Weinberg and linkage equilibrium [43,46]. To verify the accuracy of this program Frichot et al.(2014) conducted an in-depth comparison with the software ADMIXTURE using simulatedand empirical datasets and found concordant results across trials, while sNMF outperformedADMIXTURE when population inbreeding (FIS) was high. For our dataset ancestry coeffi-cients (K) were estimated using sNMF to determine subpopulation membership by running 10replicates of K 2–6 using a cross-entropy criterion (CEC). To evaluate the predictive capabilityand error of the ancestry estimation algorithm, sNMF employs the CEC, which is comparableto the likelihood value implemented in the program ADMIXTURE. To select the best-sup-ported ancestry coefficient, the lowest CEC value was represented by the K value (K = 3). Theancestry coefficient plot (Fig. 2C) was visualized using R v.3. For information on CEC values,as well as results obtained between sNMF and STRUCTURE on a subset of the total data(SNP = 2000) see S1 Methods.

The program Treemix [47] was utilized to infer the phylogenetic relationships between sam-pled locations while accounting for ancestral admixture among populations. Specifically, Tree-mix incorporates a model to allow for population divergence in the presence of post-divergence admixture/migration (m) given that incorporation of this parameter can improvethe likelihood fit of a bifurcating phylogeny. More specifically, the m parameter represents theproportion of admixture from one population to another [48]. The resulting phylogeny isbased on a composite maximum likelihood of the local optimum tree, determined using a simi-lar approach to Felsenstein [49], with branch lengths proportional to the amount of geneticdrift that has occurred per branch.

Population genetic statistics (Table 1; Fig. 2D-2F) were generated using vcftools and calcu-lated across all SNPs per individual. The calculation of FST utilized between subpopulations[50] specifically accounts for differences in sample size and a small number of sampled individ-uals, and recent studies have shown that bi-allelic SNPs (>1000) using this approach will resultin precise FST estimates [51].

To investigate the visual similarity between genetic and geographic distance from the PCAanalysis (Fig. 2A and 2B), we conducted a test for isolation-by-distance (IBD) to see if this pat-tern meets the expectation of genetic similarity decaying with geographic distance [52] usingthe IBD program by Mantel’s test (10,000 randomizations) of linearized FST (FST/(1-FST)) andshoreline distance (km) [53]. Pairwise FST, calculated in vcftools, and distance of coastlines be-tween sampling locations in kilometers was determined using Google Earth Tools. For thisMantel test, the centroid distance between sampling locations for the Gulf-Keys subpopulation

Table 1. Population genetic summary statistics for each subpopulation.

Subpopulations n FST SNPs Mean He Mean S

North-Atlantic 9 North-Atlantic South-Atlantic 11,708 1,104 304

South-Atlantic 5 0.1012 —————— 11,708 1,090 191

Gulf-Keys 9 0.083 0.0454 11,708 1,183 309

FST, number of individuals (n), total number of SNPs, observed mean heterozygote alleles per individual/per subpopulation (He) and mean number of

singletons (S) per individual/per subpopulation.

doi:10.1371/journal.pone.0116219.t001

Seahorses of the Temperate Virginia Province

PLOS ONE | DOI:10.1371/journal.pone.0116219 January 28, 2015 6 / 12

was utilized and results indicated a non-significant correlation between geographic and geneticdistance (p = 0.4925). See S1 Methods for additional information and regression plots.

Results and Discussion

Support for northern subpopulation divergence and isolationOur results strongly support H2 over H1 with Virginia Province residents of H. erectus comingfrom a persistently breeding and isolated ancestral gene pool, rejecting the categorization of itbeing composed of seasonal migrants. The sNMF-based estimates of ancestry coefficients sup-port three distinct subpopulations with limited admixture (K = 3) (Fig. 2C): 1) the eastern Gulfof Mexico-Florida Keys (Gulf-Keys), 2) the eastern Floridian Peninsula (South-Atlantic), and3) Chesapeake Bay-New York (North-Atlantic). The K = 3 value reported in our study is con-sidered robust as it exhibited the lowest CEC value across replicate runs of all K values (K =2–6). This substructure also visually emerges from the first two principle components of thePCA from the total amount of observed genomic variation (Fig. 2B). Here, the individualsfrom north of Cape Hatteras form a tight cluster, while individuals sampled from the Gulf-Keys and South-Atlantic form a cline between the tightly clustered South-Atlantic individualsand an admixed set of Gulf-Keys individuals. Consistent with these results is the inferred popu-lation history that emerges from Treemix, which is concordant with long-term isolation of theNorth Atlantic sub-population with limited post-divergence admixture withsouthern subpopulations.

The elevated heterozygosity found in Gulf-Keys individuals (Fig. 3A) could be the result ofadmixture from un-sampled western Gulf/Caribbean individuals, which is also indicated fromthe sNMF analysis (Fig. 2C). However, this elevated heterozygosity could also be the result of alarger effective population size [54]. In contrast, the northern subpopulation shows a reductionin heterozygosity with an elevated level of singletons (Fig. 3B). This pattern indicates a possibledemographic expansion after the last glacial maximum that is consistent with the likely unsuit-able habitat in the Virginia Province during the late Pleistocene. This history of shifting habitat

Figure 3. Distribution of heterozygote and singleton genotypes. Boxplots represent the range of observed heterozygote genotypes (a) and singletongenotypes per individual/per subpopulation (b).

doi:10.1371/journal.pone.0116219.g003

Seahorses of the Temperate Virginia Province

PLOS ONE | DOI:10.1371/journal.pone.0116219 January 28, 2015 7 / 12

driven by climate change is suggested by palaeo-climatological research indicating that temper-ate environments north of Cape Hatteras were displaced southward [6,55], as well as the for-mation of the Chesapeake Bay 7.4–8.2 kya due to post-last glacial maximum sea level rise [56].

Causes of divergence and isolation of the northern subpopulationGiven the strong evidence we report for Virginia Province inhabitants of H. erectus represent-ing a persistently isolated independent subpopulation from other regional ancestral gene pools,there are several conceivable non-mutually exclusive causes of this divergence. First, seagrass isa preferred breeding habitat of H. erectus and a long gap without coastal seagrasses exists alongthe Georgia and South Carolina coastlines (roughly 600km) [57]. This barrier of unsuitablebreeding habitat between northern Florida and the Virginia Province therefore likely results inthe fish’s rarity in this area, thereby increasing genetic isolation of the northern subpopulation[58,59]. The confamilial pipefish Syngnathus floridae also shares a similar pattern of genetic di-vergence across this region of unsuitability, though the area of absence extends from the south-ern end of the Florida Peninsula to near Cape Hatteras, with the northern populationextending from North Carolina to Chesapeake Bay [60]. Secondly, long-distance migration ofH. erectus is observed to occur via Sargassum rafting driven by ocean currents [10]. Under thismode of migration, the northeastern deflection in ocean currents near Cape Hatteras towardthe Mid-Atlantic may limit the arrival of southern migrants to the Virginia province. Lastly, in-dividuals that do arrive from southern provinces may have a lower physiological tolerance totemperate conditions, reducing the chance of winter survival and also increasing the amount ofgenetic isolation. Selection correlated to the shift in macroclimate at Cape Hatteras has beenobserved in marine fishes [61], and future analysis of northern adaptation inH. erectusmayhelp decouple the potential drivers of temperate subpopulation genetic isolation. Although ourobserved patterns of genetic isolation could have emerged via a continuous isolation-by-dis-tance regime without clear breaks driving the isolation, a Mantel test resulted in a non-signifi-cant relationship between genomic and geographic distance (p = 0.4925).

Support for local seasonal migrationOur results also support local offshore migration to account for the coastal absence of H. erec-tus from Virginia Province during winter months. While extreme temperature changes influ-ence latitudinal movement of many species [1], substantial seasonal movement to and fromprovinces for H. erectus is unlikely due to its relatively weak swimming ability [62]. To avoidnearshore cold water temperatures, localized inshore-offshore migration has been reported forthe confamilial pipefish (Syngnathus fuscus), which has similar life history traits to H. erectus[63], and has also been suggested for some other species ofHippocampus [62]. As a qualitativecomparison we examined abundance records from NOAA long-term offshore trawl surveys ofS. fuscus (1972–2008;>90% 20 km off-coast; depth 10–20m) and found that they closely re-semble that of H. erectus, further supporting intercontinental shelf overwintering (For addi-tional details see S2 Methods). Regarding direct observation of this phenomena, a single recordfrom divers in 1968 documented both species “hibernating” on the shelf substrates off Long Is-land, NY [64], where they resumed swimming several minutes after being brought to the sur-face. Many fish adapt to winter temperatures by decreasing energy demands and enteringsemi-torpidity [65], though no research has been conducted on the overwintering physiologyof any Syngnathidae species. Nevertheless, localized overwintering in deeper waters may be animportant component ofH. erectus’ life history and may also account for their winter absencein estuaries of the warm-temperate eastern Floridian Peninsula (i.e., South-Atlantic).

Seahorses of the Temperate Virginia Province

PLOS ONE | DOI:10.1371/journal.pone.0116219 January 28, 2015 8 / 12

ConclusionsOverall, our results demonstrate the utility of supplementing life history information with pop-ulation genomic data when a small number of unlinked genetic loci may be insufficient to dis-cern the range of persistence in difficult-to-observe fishes. Currently, the IUCN (WorldConservation Union) Red List categorizes H. erectus as “vulnerable” based on it being com-monly collected as by-catch and sold by trawl fishermen to supply the aquarium trade [66].Our results, throughout an extensive range of this species distribution, will help inform conser-vation, as well as captive breeding efforts, by strongly supporting northern Atlantic seahorsesas a genetically distinct subpopulation. More broadly, because genomic data effectively samplesa multitude of ancestors, even with a small number of sampled individuals, the approach takenin our study shows the promise of genomic data to infer population genetic structure in rareand/or difficult to obtain species.

Supporting InformationS1 Table. Raw reads, filtered, analyzed reads, and NCBI SRA accession numbers per indi-vidual.(PDF)

S1 Methods. Additional details on sNMF CEC values, sNMF and STRUCTURE compari-son, and isolation-by-distance methods.(PDF)

S2 Methods. Comparison of NOAA Long-term bottom trawl survey of the Mid-AtlanticBight (i.e., Virginia Province) betweenHippocampus erectus and Syngnathus fuscus.(PDF)

AcknowledgmentsThank you to the Sackler Institute for Comparative Genomics, American Museum of NaturalHistory and Dr. Rob DeSalle for laboratory space and support; Stephen Harris and Tyler Jo-seph for assistance with data analysis; Clay Small (University of Oregon), N. Dunham (FloridaFish andWildlife), T. Tuckey (Virginia Institute of Marine Science), T. Gardner (AtlantisAquarium, NY), T. M. Grothues (Rutgers University, NJ) and The River Project(riverprojectnyc.org) for help with fish collections and/or providing samples.

Author ContributionsConceived and designed the experiments: JTB JWMJH. Performed the experiments: JTB. Ana-lyzed the data: JTB JDR. Contributed reagents/materials/analysis tools: JTB MJH. Wrote thepaper: JTB JW JDR MJH.

References1. Briggs PT, Waldman JR (2002) Annotated list of fishes reported from the marine waters of New York.

Northeast Nat 9: 47–80.

2. Curran M (1989) Occurrence of tropical fishes in New England waters. AAUS. 71–82.

3. Teixeira R, Musick J (2001) Reproduction and food habits of the lined seahorse,Hippocampus erectus(Teleostei: Syngnathidae) of Chesapeake Bay, Virginia. Rev Bras Biol 61: 79–90. PMID: 11340465

4. Milstein CB, Thomas DL (1976) Fishes New or Uncommon to the New Jersey Coast. Chesap Sci 17:198.

5. Briggs JC, Bowen BW (2012) A realignment of marine biogeographic provinces with particular refer-ence to fish distributions. J Biogeogr 39: 12–30. doi: 10.1111/j.1365-2699.2011.02613.x

Seahorses of the Temperate Virginia Province

PLOS ONE | DOI:10.1371/journal.pone.0116219 January 28, 2015 9 / 12

6. McCartney MA, Burton ML, Lima TG (2013) Mitochondrial DNA differentiation between populations ofblack sea bass (Centropristis striata) across Cape Hatteras, North Carolina (USA). J Biogeogr 40:1386–1398. doi: 10.1111/jbi.12103

7. McBride RS (2014) Managing a Marine Stock Portfolio: Stock Identification, Structure, and Manage-ment of 25 Fishery Species along the Atlantic Coast of the United States. North Am J Fish Manag 34:710–734. doi: 10.1080/02755947.2014.902408

8. Grosberg C, Cunningham C (2001) Genetic structure in the sea: from populations to communities. In:Bertness M, Gaines S, Hay M, editors. Marine Community Ecology. Sinauer Associates. pp. 61–84.

9. Fish MP and Mowbray MH (1970) Sounds of western North Atlantic fishes. The Johns Hopkins Univer-sity Press.

10. Casazza TL, Ross SW (2008) Fishes associated with pelagic Sargassum and open water lacking Sar-gassum in the Gulf Stream off North Carolina. Fish Bull 106: 348–363.

11. Able K, Fahay M (1998) Ecology of Estuarine Fishes: Temperate Waters of the Western North Atlantic.John Hopkins University Press.

12. Howell P, Auster PJ (2012) Phase Shift in an Estuarine Finfish Community Associated with WarmingTemperatures. Mar Coast Fish 4: 481–495. doi: 10.1080/19425120.2012.685144

13. LoweWH, Allendorf FW (2010) What can genetics tell us about population connectivity? Mol Ecol 19:3038–3051. doi: 10.1111/j.1365-294X.2010.04688.x PMID: 20618697

14. Hare MP, Nunney L, Schwartz MK, Ruzzante DE, Burford M, et al. (2011) Understanding and estimat-ing effective population size for practical application in marine species management. Conserv Biol 25:438–449. doi: 10.1111/j.1523-1739.2010.01637.x PMID: 21284731

15. Pringle JM, Wares JP (2007) Going against the flow: Maintenance of alongshore variation in allele fre-quency in a coastal ocean. Mar Ecol Prog Ser 335: 69–84. doi: 10.3354/meps335069

16. Martinez-Solano I, Gonzalez EG (2008) Patterns of gene flow and source-sink dynamics in high altitudepopulations of the common toad Bufo bufo (Anura: Bufonidae). Biol J Linn Soc 95: 824–839. doi: 10.1111/j.1095-8312.2008.01098.x

17. Mobley KB, Small CM, Jones AG (2011) The genetics and genomics of Syngnathidae: pipefishes, sea-horses and seadragons. J Fish Biol 78: 1624–1646. doi: 10.1111/j.1095-8649.2011.02967.x PMID:21651520

18. Teske PR, Cherry MI, Matthee CA (2003) Population genetics of the endangered Knysna seahorse,Hippocampus capensis. Mol Ecol 12: 1703–1715. doi: 10.1046/j.1365-294X.2003.01852.x PMID:12803625

19. Teske P, Hamilton H, Palsbøll P, Choo C, Gabr H, et al. (2005) Molecular evidence for long-distancecolonization in an Indo-Pacific seahorse lineage. Mar Ecol Prog Ser 286: 249–260.

20. Lourie SA, Green DM, Vincent ACJ (2005) Dispersal, habitat differences, and comparative phylogeo-graphy of Southeast Asian seahorses (Syngnathidae: Hippocampus). Mol Ecol 14: 1073–1094. doi:10.1111/j.1365-294X.2005.02464.x PMID: 15773937

21. Lourie SA, Vincent ACJ (2004) A marine fish followsWallace’s Line: the phylogeography of the three-spot seahorse (Hippocampus trimaculatus, Syngnathidae, Teleostei) in Southeast Asia. J Biogeogr31: 1975–1985. doi: 10.1111/j.1365-2699.2004.01153.x

22. Boehm JT, Woodall L, Teske PR, Lourie SA., Baldwin C, et al. (2013) Marine dispersal and barriersdrive Atlantic seahorse diversification. J Biogeogr 40: 1839–1849. doi: 10.1111/jbi.12127 PMID:17320419

23. Sousa V, Hey J (2013) Understanding the origin of species with genome-scale data: modelling geneflow. Nat Rev Genet 14: 404–414. doi: 10.1038/nrg3446 PMID: 23657479

24. Corander J, Majander KK, Cheng L, Merilä J (2013) High degree of cryptic population differentiation inthe Baltic Sea herring Clupea harengus. Mol Ecol 22: 2931–2940. doi: 10.1111/mec.12174 PMID:23294045

25. Hohenlohe PA, Amish SJ, Catchen JM, Allendorf FW, Luikart G (2011) Next-generation RAD sequenc-ing identifies thousands of SNPs for assessing hybridization between rainbow and westslope cutthroattrout. Mol Ecol Resour 11 Suppl 1: 117–122. doi: 10.1111/j.1755-0998.2010.02967.x PMID: 21429168

26. Hohenlohe PA, Bassham S, Etter PD, Stiffler N, Johnson E, et al. (2010) Population genomics of paral-lel adaptation in threespine stickleback using sequenced RAD tags. PLoS Genet 6: e1000862. doi: 10.1371/journal.pgen.1000862 PMID: 20195501

27. Deagle BE, Jones FC, Absher DM, Kingsley DM, Reimchen TE (2013) Phylogeography and adaptationgenetics of stickleback from the Haida Gwaii archipelago revealed using genome-wide single nucleo-tide polymorphism genotyping. Mol Ecol 22: 1917–1932. doi: 10.1111/mec.12215 PMID: 23452150

Seahorses of the Temperate Virginia Province

PLOS ONE | DOI:10.1371/journal.pone.0116219 January 28, 2015 10 / 12

28. Catchen J, Bassham S, Wilson T, Currey M, O’Brien C, et al. (2013) The population structure and re-cent colonization history of Oregon threespine stickleback determined using restriction-site associatedDNA-sequencing. Mol Ecol 22: 2864–2883. doi: 10.1111/mec.12330 PMID: 23718143

29. Wagner CE, Keller I, Wittwer S, Selz OM, Mwaiko S, et al. (2013) Genome-wide RAD sequence dataprovide unprecedented resolution of species boundaries and relationships in the Lake Victoria cichlidadaptive radiation. Mol Ecol 22: 787–798. doi: 10.1111/mec.12023 PMID: 23057853

30. Li S, JakobssonM (2012) Estimating demographic parameters from large-scale population genomicdata using Approximate Bayesian Computation. BMCGenet 13: 22. doi: 10.1186/1471-2156-13-22PMID: 22453034

31. Felsenstein J (2006) Accuracy of coalescent likelihood estimates: do we need more sites, more se-quences, or more loci? Mol Biol Evol 23: 691–700. doi: 10.1093/molbev/msj079 PMID: 16364968

32. Robinson J, Bunnefeld L, Hearn J, Stone G, Hickerson MJ (2014) ABC inference of multi-population di-vergence with admixture from un-phased population genomic data. Mol Ecol 18: 4458–4471. doi: 10.1111/mec.12881 PMID: 25113024

33. Gronau I, Hubisz MJ, Gulko B, Danko CG, Siepel A (2011) Bayesian inference of ancient human de-mography from individual genome sequences. Nat Genet 43: 1031–1034. doi: 10.1038/ng.937 PMID:21926973

34. Lohse K, Harrison RJ, Barton NH (2011) A general method for calculating likelihoods under the coales-cent process. Genetics 189: 977–987. doi: 10.1038/ng.937 doi: 10.1534/genetics.111.129569 PMID:21900266

35. Hearn J, Stone GN, Bunnefeld L, Nicholls JA, Barton NH, et al. (2014) Likelihood-based inference ofpopulation history from low coverage de novo genome assemblies. Mol Ecol 23: 198–211. doi: 10.1111/mec.12578 PMID: 24188568

36. Baird NA, Etter PD, Atwood TS, Currey MC, Shiver AL, et al. (2008) Rapid SNP discovery and geneticmapping using sequenced RADmarkers. PLoS One 3: e3376. doi: 10.1371/journal.pone.0003376PMID: 18852878

37. Lozier JD (2014) Revisiting comparisons of genetic diversity in stable and declining species: assessinggenome-wide polymorphism in North American bumble bees using RAD sequencing. Mol Ecol 23:788–801. doi: 10.1111/mec.12636 PMID: 24351120

38. Langmead B, Trapnell C, Pop M, Salzberg SL (2009) Ultrafast and memory-efficient alignment of shortDNA sequences to the human genome. Genome Biol 10: R25. doi: 10.1186/gb-2009-10-3-r25 PMID:19261174

39. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, et al. (2009) The Sequence Alignment/Map formatand SAMtools. Bioinformatics 25: 2078–2079. doi: 10.1093/bioinformatics/btp352 PMID: 19505943

40. Danecek P, Auton A, Abecasis G, Albers CA, Banks E, et al. (2011) The variant call format andVCFtools. Bioinformatics 27: 2156–2158. doi: 10.1093/bioinformatics/btr330

41. R Development Core Team (2004) R: A Language and Environment for Statistical Computing.

42. Patterson N, Price AL, Reich D (2006) Population structure and eigenanalysis. PLoS Genet 2: e190.doi: 10.1371/journal.pgen.0020190

43. Frichot E, Mathieu F, Trouillon T, Bouchard G, François O (2014) Fast and efficient estimation of indi-vidual ancestry coefficients. Genetics 196: 973–983. doi: 10.1534/genetics.113.160572 PMID:24496008

44. Pritchard JK, Stephens M, Donnelly P (2000) Inference of Population Structure Using Multilocus Geno-type Data. Genetics 155: 945–959.

45. Alexander DH, Novembre J, Lange K (2009) Fast model-based estimation of ancestry in unrelated indi-viduals. Genome Res 19: 1655–1664. doi: 10.1101/gr.094052.109 PMID: 19648217

46. Harris SE, O’Neill RJ, Munshi-South J (2014) Transcriptome resources for the white-footed mouse(Peromyscus leucopus): new genomic tools for investigating ecologically divergent urban and ruralpopulations. Mol Ecol Resour: Early View. doi: 10.1111/1755-0998.12301

47. Pickrell JK, Pritchard JK (2012) Inference of population splits and mixtures from genome-wide allele fre-quency data. PLOSGenetics 8: 28. doi: 10.1371/journal.pgen.1002967 PMID: 23166502

48. Gompert Z, Lucas LK, Buerkle CA, Forister ML, Fordyce JA, et al. (2014) Admixture and the organiza-tion of genetic diversity in a butterfly species complex revealed through common and rare genetic vari-ants. Mol Ecol 23: 4555–4573. doi: 10.1111/mec.12811 PMID: 24866941

49. Felsenstein J (1981) Evolutionary trees from gene frequencies and quantitative characters: findingmaximum likelihood estimates. Evolution 35: 1229–1242.

50. Weir BS, Cockerhan CC (1984) Estimation of gene flow from F-statistics. Evolution 38: 1358–1370.

Seahorses of the Temperate Virginia Province

PLOS ONE | DOI:10.1371/journal.pone.0116219 January 28, 2015 11 / 12

51. Willing E-M, Dreyer C, van Oosterhout C (2012) Estimates of genetic differentiation measured by F(ST)do not necessarily require large sample sizes when using many SNPmarkers. PLoS One 7: e42649.doi: 10.1371/journal.pone.0042649 PMID: 22905157

52. Novembre J, Johnson T, Bryc K, Kutalik Z, Boyko AR, et al. (2008) Genes mirror geography within Eu-rope. Nature 456: 98–101. doi: 10.1038/nature07331 PMID: 18758442

53. Jensen J, AJ B, Kelley S (2005) Isolation by distance, web service. BMCGenet 6: http://ibdws.sdsu.edu/.

54. Gazave E, Chang D, Clark AG, Keinan A (2013) Population growth inflates the per-individual number ofdeleterious mutations and reduces their mean effect. Genetics 195: 969–978. doi: 10.1534/genetics.113.153973 PMID: 23979573

55. Cronin TM, Szabo BJ, Ager TA, Hazel JE, Owens JP (1981) Quaternary climates and sea levels of theU.S. Atlantic coastal plain. Science 211: 233–240. doi: 10.1126/science.211.4479.233 PMID:17748008

56. Bratton JF, Colman SM, Thieler ER, Seal RR (2002) Birth of the modern Chesapeake Bay estuary be-tween 7.4 and 8.2 ka and implications for global sea-level rise. Geo-Marine Lett 22: 188–197. doi: 10.1007/s00367-002-0112-z

57. Short F, Carruthers T, DennisonW, Waycott M (2007) Global seagrass distribution and diversity: A bio-regional model. J Exp Mar Bio Ecol 350: 3–20. doi: 10.1016/j.jembe.2007.06.012

58. Lourie SA, Foster SJ, Cooper EWT, Vincent ACJ (2004) A guide to the identification of seahorses. Proj-ect Seahorse and TRAFFIC North America.

59. Wenner CA, Sedberry GR (1989) Species composition, distribution, and relative abundance of fishes inthe coastal habitat off the southeastern United States. NOAA Technical Report NMFS 79: 1–78.

60. Mobley KB, Small CM, Jue NK, Jones AG (2010) Population structure of the dusky pipefish (Syng-nathus floridae) from the Atlantic and Gulf of Mexico, as revealed by mitochondrial DNA and microsatel-lite analyses. J Biogeogr 37: 1363–1377. doi: 10.1111/j.1365-2699.2010.02288.x

61. Hice LA, Duffy TA, Munch SB, Conover DO (2012) Spatial scale and divergent patterns of variation inadapted traits in the ocean. Ecol Lett 15: 568–575. doi: 10.1111/j.1461-0248.2012.01769.x PMID:22462779

62. Foster SJ, Vincent ACJ (2004) Life history and ecology of seahorses: implications for conservation andmanagement. J Fish Biol 65: 1–61. doi: 10.1111/j.0022-1112.2004.00429.x

63. Lazzari MA, Able KW (1990) Northern pipefish, Syngnathus fuscus, occurrences over the Mid-AtlanticBight continental shelf: evidence of seasonal migration. Environ Biol Fishes 27: 177–185. doi: 10.1007/BF00001671.x

64. Wicklund R, Wilk S, Ogren L (1968) Observations on wintering locations of the northern pipefish andspotted seahorse. Underw Nat 5: 26–28.

65. Ultsch G (1989) Ecology and physiology of hibernation and overwintering among freshwater fishes, tur-tles, and snakes. Biol Rev 64: 435–515. doi: 10.1111/j.1469-185X.1989.tb00683.x

66. Dias TL, Rosa I, Baum JK (2002) Threatened fishes of the world: Hippocampus erectus Perry, 1810(Syngnathidae). Environ Biol Fishes 65: 326. doi: 10.1023/A:1020539222587

67. Tyberghein L, Verbruggen H, Pauly K, Troupin C, Mineur F, et al. (2012) Bio-ORACLE: a global envi-ronmental dataset for marine species distribution modelling. Glob Ecol Biogeogr 21: 272–281. doi: 10.1111/j.1466-8238.2011.00656.x

Seahorses of the Temperate Virginia Province

PLOS ONE | DOI:10.1371/journal.pone.0116219 January 28, 2015 12 / 12